Elucidation and Decisional Risk in a Multi Criteria Decision based on a Choquet Integral Aggregation—A Cybernetic Framework

The authors are developing multi criteria Decision-Making Support Systems (DMSS) for project teams in charge of selecting a technical solution among alternatives. They propose a cybernetic framework to emphasize the link between decision-making and knowledge management processes in such projects. These DMSS's rely on the tracking of the accompanying knowledge production of long-term decisional processes by a collective with many actors. Based on knowledge production management, this paper explains how to design decisional risk evaluation, monitoring and control aids and traceability functions for strategic choices and logical argumentation. The DMSS is seen as a recommender system for the project manager. Each possible solution involved in the decision-making process is evaluated by means of a set of criteria. The evaluation results from an interpretation of the knowledge items in terms of satisfaction scores of the solutions according to the considered criteria. Aggregating these partial scores provides a ranking of all the possible solutions by order of preference. As criteria are sometimes interacting, the aggregation has to be based on adapted operators, i.e. Choquet integrals. Evaluating possible solutions by the knowledge contained in the knowledge base opens the way to automating the argumentation of the project team's decisions: the argumentation principle underlying this approach is based naturally on coupling a knowledge dynamical management system (KDMS) with the DMSS. The DMSS also evaluates the decisional risk that reflects the eventuality of a wrong selection due to the insufficiency of available knowledge at a given time in order to adopt a reliable solution. Decisional risk assessment corresponds to sensitivity analyses. These analyses are then exploited to control the decisional risk in time: they enable to identify the crucial information points for which additional and deeper investigations would be of great interest to improve the stability of the selection in the future. The knowledge management of a collective project is represented as a control loop: the KDMS is the actuator, the risk accompanying the decision is the controlled variable and is strongly linked to the entropy of the knowledge base managed by the KDMS. Each of the three phases—Intelligence, Design, Choice—of the decision-making process is identified to a function of the control loop: actuator, process and regulator. This cybernetic framework for decision has its origin in knowledge management activities for a great-scale project—the EtLD project of the French Atomic Commission (CEA) that concerns the management of high-level long-life radioactive waste in France.